结合新型模糊支持向量机和证据理论的多传感器水质数据融合

在多传感器水质数据融合领域,证据理论是有效的数据融合方法之一,但基本概率分配一般不易确定,从而使数据融合能力难以有效发挥。支持向量机是统计学习理论之上的高级分类算法,具有普适性和全局优化等特点,但输出的基本概率分配有待进一步提高。提出了一种基于证据理论和新型模糊支持向量机相结合的数据融合方法,通过建立基于分类超平面距离的模糊隶属度,训练模糊支持向量机提高传统支持向量机的基本概率分配,并结合证据理论进行海河水质数据融合。通过证据理论分别结合支持向量机和模糊综合评价法与上述方法进行对比实验,经精度、平均绝对百分误差、均方根误差等指标验证,精度提高10. 5%,表明所提方法是一种可靠的多传感器的水质融合方法,较其他方法具有更高的融合精度。

Multi-sensor water quality data fusion combining novel fuzzy SVM and D-S evidence algorithm

In the field of multi-sensor water quality data fusion,Dempster-Shafer(D-S) evidence theory is one of the effective methods,but the basic probability assignment(BPA) is not easy to determine,which limits data fusion ability of D-S to some extent. Support vector machine( SVM) is an advanced classification algorithm based on statistical learning theory and it has the characteristics of universality and global optimization. However,its indirect BPA output to various water grade needs to be further improved. In this paper,a data fusion model based on D-S evidence algorithm and new fuzzy SVM(FSVM) is proposed.Through establishing the membership of distance to hyperplane,FSVM can improve the classification accuracy of traditional SVM,thus providing more accurate BPA to D-S evidence algorithm. To prove the performance of model,D-S evidence is combined with SVM,FSVM and fuzzy comprehensive evaluation method(FSE) to carry out fusion experiments respectively,aiming at Haihe multi-sensor data fusion. The accuracy,mean absolute percentage error(MAPE),root mean square error(RMSE) are imported to calculate errors and the result shows that the accuracy is enhanced by 10. 5%,so the model is a more reliable water quality fusion method with higher fusion accuracy than other methods.